Simple Procedure for Seismic Analysis of LiquidStorage Tanks
Praveen K. Malhotra, Senior Res. Scientist
Factory Mutual Research. Norwood. MA, LISA
Thomas Wenk, Civil Eng.
Swiss Federal Institute oí Technology. Zurich, Switzerland
Martin Wieland, Dr
Elcciriiw;ilt Engineering Lid. Zurich. Switzerland
Summary unanchored tanks supported on rigid
foundations were therefore studied
This paper provides the theoretical background of a simplified seismic design
[ 151. It was shown that base uplifting
reduces the hydrodynamic forces in
procedure for cylindrical groundsupported tanks The procedure takes into ac
the tank, but increasessignificantly the
count impulsive and convective (sloshing) actions of the liquid in flexible steel or
axial compressive stress in the tank
concrete tanks fixed to rigid foundations. Seismic responses  base shear. over
wall.
turning moment, and sloshing wave height  are calculated by using the site re
sponse spectra and pcrforming a few simple calculations An example is present Further studies [16,171 showed that
ed to illustrate the procedure. and a comparison is made with the detailed modal base uplifting in tanks supported
analysis procedure. The simplified procedure has been adopted in Eurocode 8. directly on flexible soil foundations
does not lead to a significant increase
L
in the axial compressive stress in the
Introduction tions It was shown that a part of the
tank wall, but may lead to large foun
liquid moves in longperiod sloshing
dation penetrations and several cycles
motion, while the rest moves rigidly
Largecapacity groundsupported cylin of large plastic rotations at the plate
with the tank wall. The latter part of
drical tanks are used to store a variety boundary. Flexibly supported unan
the liquid  also known as the impul
of liquids. e.g. water for drinking and chored tanks are therefore less prone
sive liquid  experiences the same
fire fighting. petroleum, chemicals. to elephantfoot buckling damage, but
acceleration as the ground and con
and liquefied natural gas. Satisfactory more prone to uneven settlement of
tributes predominantly to the base
performance of tanks during strong the foundation and fatigue rupture at
shear and overturning moment. The
ground shaking is crucial for modern the plateshell junction.
sloshing liquid determines the height
facilities. Tanks that were inadequately
designed or dctailcd have suffered
of the freesurface waves, and hence In addition to the above studies, nu
the freeboard requirement. merous other experimental and nu
extensive damage during past earth
quakes [l71. merical studies have provided valuable
It was shown later [lo121 that the
insight into the seismic behaviour of
Earthquake damage to steel storage flexibility of the tank wall may cause
the impulsive liquid to experience tanks [18271. This paper deals only
tanks can take several forms. Large ax with the elastic analysis of fully an
accelerations that are several times
ial compressive stresses due to beam chored, rigidly supported tanks The
greater than the peak ground acceler
like hending «( thc tank wall can cause effects of foundation flexibility and
ation. Thus the base shear and over
"elephantfoot" buckling of the wall base uplifting on the tank response
(Fig. I). Sloshing liquid can damage
turning moment calculated by assum
may be found elsewhere [13171.
ing the tank to be rigid can be non
the roof and the top of tank wall
conservative. Tanks supported on flex
(Fig. 2). High stresses in the vicinity of
ible foundations, through rigid base
poorly detailed hase anchors can rup
mats, experience base translation and
turc the tank wall. Base shear can
rocking, resulting in longer impulsive
overcome friction causing the tank to
periods and generally greater effective
slide. Base uplifting in unanchored or
damping. These changes may affect
partially anchored tanks can damage
the piping connections that arc inca the impulsive response significantly
[13.14]. The convective (or sloshing)
pable of accommodating vertical dis
response is practically insensitive to
placements. rupture the plateshell
junction due to excessive joint stresses, both the tank wall and the foundation
flexibility due to its long period of
and cause uneven settlement of the
foundation. oscillation.
Tanks analysed in the above studies
Initial analytical studics (X. U] dealt
with the hydrodynamics of liquids in
were assumed to be completely an
chored at their base. In practice, com
rigid tanks resting on rigid founda
plete base anchorage is not always fea
sible or economical. As a result, many
tanks are either unanchored or only
Pcerreviewed by intcmilli«nal ex partially anchored at their hase. The
perts end accepicd fur publication effects of hase uplifting on the seismic
hy IABSF Puhlicaiiiins Committee response of partially anchored and
adjusting the impulsive and convec
tive heights to account for the over
turning effect of the higher modes
 generalising the impulsive period
formula so that it can be applied to
steel as well as concrete tanks of
various wall thikknesses.
The impulsive and convective respons
es are combined by taking their nu
merical sum rather than their root
meansquare value.
Model Properties
The natural periods of the impulsive
(TimP) and the convective (T,,,,)
responses are
Fig. 2 Sloshing damage ro upper shell oftunk ( ~ t u n q L + P W , , ~qc~tq¿l<ir~
~~~ <I, B W ~ < , I ~ ? J
TImp= c; 7HziP
h/rx E
(1) J
L2 = c,& (2)
Method of Dynamic Analysis and first convective modes are consid where h is the equivalent uniform
ered satisfactory in most cases. There thickness of the tank wall. p the mass
The dynamic analysis of a liquidfilled is. however. some merit in slightly ad density of liquid, and E the modulus of
tank may be carried out using the con justing the modal properties of these elasticity of the tank material. The co
cept of generalised singledegreeof two modes to account for the entire efficients C, and C, are obtained from
freedom (SDOF) systems representing liquid mass in the tank. Fig. 4 or Table 1. The coefficient Ci is
the impulsive and convective modes of dimensionless, while C, is expressed in
vibration of the tankliquid system. d&. For tanks with nonuniform wall
For practical applications, only the first Simple Procedure for Seismic thickness, h may he calculated by tak
few modes of vibration need to be con Analysis ing a weighted average over the wetted
sidered in the analysis (Fig.3). The height of the tank wall, assigning the
mass, height and natural period of highest weight near the base of the
The procedure presented here is based
each SDOF system are obtained by the on the work of Veletsos and cowork tank where the strain is maximal.
methods described in [lo141. For a ers [lo, 12. 141 with certain modifica
given earthquake ground motion. the tions that make the procedure simple.
response of various SDOF systems yet accurate, and more generally ap

‘1) r L  ~ ~ ~  . ~
may be calculated independently and plicable. Specifically. these modifica ~
then combined to give,the net base tions include
shear and overturning moment.
representing the tankliquid system
by the first impulsive and first con
vective modes only
 combining the higher impulsive modal
mass with the first impulsive mode
and the higher convective modal
mass with the first convective mode
~
H/r C, C . [siVz] rn,/m, mchn, h,/H h,/H h,7H h,’/H
Fig. 3: Liquidfilled rank modelled by gen ~
eralised singledegreeo~freedom ,s.vsfenu 0.3 9.28 2.09 0.176 o.xz4 0.400 0.521 2.640 3.414
0.5 7.74 1.74 0.3w (J. 7íhl 0.Joo 0.543 1.460 1.517
For most tanks (0.3 < H/r < 3, where 0.7 6.97 1.60 0.414 0586 0.401 0.571 ].MI9 1.01 1
His the height of water in the tank and
1.0 6.36 1.52 0.548 0.4.52 0.419 0.616 0.721 0.785
r the tank radius), the first impulsive
and first convective modes together 1.5 6.06 1.48 0.686 0.311 0.439 0.6W 0.555 0.734
account for 8598% of the total liquid 2.0 6.21 1.48 0.763 (1.237 0,448 0.751 0.500 0.764
mass in the tank. The remaining mass 2.5 6.56 1.48 0.810 o. 190 0.452 0.794 0.480 0.796
of the liquid vibrates primarily in high 3.0 7.03 1.48 0.wz 0.158 0.453 0.825 0.472 0.82s
er impulsive modes for tall tanks
(H/r > 1), and higher convective modes Tahle I : Recommended design wIue.s for rhe,ñrsr impulsive and convective modes ofvibra
for broad tanks (H/r 5 1). The results Non as a firncrion ofrhe rank heiglirtoradiusratio (Wr). All coefficients are based on an
obtained using only the first impulsive exacr model oftbe tunkliqiiid sysiem [ / O , 12. 141.
I I I
n.n
O.6
0.4
(6)
11.2
on the hydrodynamic pressure on the responses were calculated first, using
HIR
tank wall as well as that on the base the rootmeansquare rule, then nu
Fig. 5: Impulsive and convective masses CIS
fractions of the total liquid mass in rhe tank
plate. it is given by Eq. (5). where the merically added to give the overall re
heights h,' and h, ' are obtained from sponse. The base shear, for example,
Fig. 6 or Table 1. was obtained using Eq. (7). where Q,'
and Q<'are the base shear values for
The impulsive and convective masses if the tank is supported on a ring foun
the first impulsive and first convective
(mi and m,)are obtained from Fig. 5 or dation, Mshould be used to design the
modes, respectively. The response
Table I as fractions of the total liquid tank wall, base anchors and the foun
spectra for the site are the same as
mass (mt). dation. If the tank is supported on a
those used in the given example
mat foundation. M should be used to
(Fi8 7).
design the tank wall and anchors only,
Seism ie Responses
while M' should he used to design the The results (Tubh2) show that the
The total base shear is given by foundation. values of base shear and moment ob
tained from the proposed procedure
The vertical displacement of the liquid
Q = (ini + FJI" + m, I x S,JTrn,,)+ ni,S,íT.,,d (3) were 210% higher than those from
surface due to sloshing ( d ) is given by
Eq. (6). where g is the acceleration due
where rnW is the mass of tank wall. rn,
the mass of tank roof, Sc(Timp)the im
to gravity. /. 7
pulsive spectral acceleration (obtained
from a 2% damped elastic response Comparison with Detailed
spectrum for steel and prestressed
concrete t a n k s or a 5% damped elastic
Modal Analysis
response spectrum for concrete tanks),
Three steel tanks were selected for
and S,(T,,) the convective spectral
comparing the results obtained from
acceleration (obtained from a 0.5%
the proposed procedure with those
damped elastic response spectrum).
from a detailed modal analysis. Three
The overturning moment above the impulsive and three convective modes
base plate, in combination with ordi were used in the detailed analysis. The
nary beam theory, leads to the axial modal analysis results were calculated
stress at the base of the tank wall. The using a combination of rootmean Period. T (I)
net overturning moment immediately square and algebraicsum rules. The Fig. 7: Elastic design response spectrn for
above the base plate (M) is given by net impulsive and the net convective 0.5% ond 2% damping
Eq. (4). where h, and h,. are the heights
of the centroids of the impulsive and
convective hydrodynamic wall pres
sures (Fig.6. Table I). and h, and h,
are the heights of the centres of gravity r[m] H[m] hh Q lMNl M [MNm] M'[MNm] D [cm]
of the tank wall and roof. respectively.
15 7.5 0.WI 15.7 (14.Y) 49.3 (44.6) 167 (164) 57 (51)
The overturning moment immediately 15 15 O.(X)l 53.6 (52.4) 346 (334) 577 (557) 75 (h6)
below the base plate (M') is dependent 7.5 15 0.001 18.3 (16.5) 127 (123) 140 (136) 79 (67)
Table 2: Comparrron of results from proposed procedure with those from detailed unalysis
(values in parenthrses are from [he modal analysis)
3
Tank contents Importance factor f n) for
Class 1 Class2 Class3
Drinking water, nontoxic nonflammable chemicals 1.2 1.o 0.8
Firefighting water, nonvolatile toxic chemicals 1.4 1.2 1.0
o 1l.S I 1.5 2 2,s 3 lowly flammable petrochemicals
HIR
Volatile toxic chemicals, explosive and highly 1.6 1.4 1.2
Fig. 6: Impulsive and r o n v k i v e heights as flammable liquids
fractions of the hcight of the liquid in the
rank Table2 Imporlanccfa(.lor /n)for tanks according to Eiuocode 8 1281
the detailed modal analysis The values modal analysis. The results of the pro Design According to
of slashing wave height obtained from posed procedure are therefore con Eurocode 8
the proposed procedure were 1218% servative hut close to those from the
higher than those from the detailed detailed modal analysis The presented simple procedure was
used in Eurocode 8 1281 and integrated
Example in its limit state design concept. The
serviceability and ultimate limit states
A steel tank with a radius I of 10 m and total height of 9.6 m is fully anchored to have to he verified. The specification
a concrete mat foundation. The tank is filled with water to a height H of of the corresponding seismic actions is
8 m (rUr = 0.8). The total mass of water in the tank (ml)is 2.51 x 1íY kg. The left to the national authorities. The
tank wall is made of four courses,each 2.4 m high. The lower two courses are 1 level of seismic protection is estah
mthick and the upper two courses 0.8 cm thick. The total mass of the tank wall lished based on the risk to life and the
(m,) is 43 x 10' kg, and the height of its centre of gravity (h,) is 4.53 m. The economic and environmental conse
mass of the tank roof (m,)is 25 x 10' kg and the height of its centre of gravity quences. This reliability differentiation
(h,) is 9.6 m. The 0.5% and 2% damped elastic response spectra for the site are is achieved by adjusting the return pe
shown in Fig. 7. riod of the design seismic event. Three
tank reliability classes are defined cor
M&lRvpe&ea responding to situations with high
F i t , the equivalent uniform thickness of the tank wall is calculated by (Class I). medium (Class 2) and low
the weighted average method. Using weights equal to the distance from the (Class 3) risk. Depending on the tank
liquid surface contents an importance factor (E) is
assigned to each of the three classes
a01x24x6.8+0.01x 2 4 x4.4 +O.O08x 2 4 x 2 + 0.008 x 0.8x0.4 = o,oo968 (Table 7).
h=
24~&8+24~4.4+24~2+0.8~0.4 The seismic action effects have to be
For steel, E = 2 x 10" N/m'. For water, p = loo0 kg/m'. For H/r = 0.8,Ci = 6.77 multiplied by the selected importance
and C, = 1.57 slmoJ(Table I). Hence, from Eqs (1) and (2). factor. For the reference case (y, = 1).
the recommended return periods of
8 ~ 4 6 % the design seismic event are 475 years
Tmp= 6.77 x =0.123 s
4!ziGEE x J2 x 10" for the ultimate limit state and 5070
years for the serviceability limit state.
In the case of the largest importance
T = 1.57xfi =4.96 s factor (y, = 1.6). the return period of
the design event for the ultimate limit
For H/r = 0.8,mi/mi= 0.459 and m,/ml= 0.541 (Table 1).Hence, state is about 20" years. According to
Eurocode 8. the analysis has to assume
i?lj =a459X25i X I 0 6 = 1.15X 106 kg linear elastic hehaviour. allowing only
for localised nonlinear phenomena
4=a54iX25iX106 1 . 3 6 lo6
~ kg without affecting the global response.
and to include the hydrodynamic re
Also from Table 1, hi/H = 0.404,h,/H = 0.583, h//H = 0.891, h,'/H = 0.954. sponse of the fluid. Particularly. it
Hence, hi = 3.23 m, h, 2 4.66 m, h;= 7.13 m. and h,' = 7.63 m. should account for the convective and
impulsive components of fluid motion
Seismic Responses as well as the tank shell deformation
The impulsive spectral acceleration for Tjv = 0.123 s, obtained from the 2% due to hydrodynamic pressure and
damped elastic response spectrum (Fig.4). is S,(T,) = 0.87413. The convective interaction effects with the impulsive
spectral acceleration for T,,, = 4.96 s, obtained from the 0.5% damped component. The proposed procedure
response spectrum in Fig. 4. is &(T, J = 0.07 g. satisfies these principles in a simple
and efficient way for the design of
The base shear obtained from Eq. (3) is fixedbase cylindrical tanks.
4 ~ + 1.36 x lo6 x 0.07 x 9.81 = 11 MN
Q = (1.15+(1M3+0.O25) x 106 ~ 0 . 8 7 9.81
Future Research Needs
The overturning moment above the base plate, obtained from Eq. (4), is
In regions of strong ground shaking.
A4 4 . 1 5 3.23+0.O43
~ x 4.53+0.025 x 9.6) x lo6 x 0.874 x9.81+ 1.36X lo6 x 4.66 it is sometimes impractical to design
=40 MNm tanks for forces obtained from elastic
and the overturning moment below the base plate, obtained from Eq. (5). is (no damage) response analysis. Elastic
forces are so large that they are arbi
~ = ( 1 . 1 5 ~ 7 . 1 3 + ~ 0 4 3 x 4 . 5 3 + 0 . M 5 ~ 9 . 6 ) ~ 1 0 6 x 0 . 8 7 4 x 9 . 8 1 + 1 . 3 6 ~ 1 0 ~ ~ 7 . 6 3 ~ 0 trarily
. 0 7 ~ 9 .reduced
81 by factors of 3 or more
81 MNm to obtain the design forces. When suh
jected to strong shaking, tanks there
The maximum vertical displacement of the liquid surface due to sloshing, fore respond in a nonlinear fashion
obtained from Eq. (6). is and experience some damage. How
d=lOx0.07=0.7 m ever, no generally acceptable methods
exist to perform a nonlinear seismic
analysis of tanks. Therefore, the dam (81 JACOBSEN. L. S Impuhivr fi.ydrudynluiiirr (201 MANOS, G. C.: CLOUGH, R. W. Funher
age sustained by tanks subjected to of fluid inside a cylindrical rank and of puid srudy of the eorthquokr responw of a broad
surrounding a cylinddwl pier. Bulletin oi the cylindrical liquid.~rorage tank modcl. Report
ground motions of different intensities
Seismological Society of America. Vol. 39. No. 3. EERC 8207. University oi California, Berke
cannot be quantified easily. There is a 1949. pp. 189203. ley. 1982.
need for practical methods of nonlin
ear analysis and design of liquidstor [ Y ] HOUSNER. G. W. The dynamic bchnvior of' (211 AULI, W.: FISCHER, E D.: RAMMERS
water mnks Bulletin of the Scismological Soci TORFER. F.G. Uplijiing of eanhquake.looded
age tanks. etyof America, Vol. 53. No. 2. IYh3, pp. 3131387. liquidlil(cd tanks Proceedings of the Pressure
Unlike ductile building systems. tanks [lo] VELETSOS, A. S,YANG. J. Y.Earthquake Vessels and Piping Conference, New Orleans,
lack a mechanism to dissipate large response of liquid rtorogc 1ank.r. Proccedings of
vol. 98, NO. 7, i w , p p . 7185,
amounts of seismic energy in a ductile the Second Engineering Mechanics Specialty [22] LEON. G. S.: KAUSEL, E. A. M. Srisrnic
manner. Methods of improving the Conference. ASCE, Raleigh. iY77. pp. 124. analysis offluid storagc tanks Journal of Struc
seismic performance of tanks by in tural Engineering. ASCE. New York. Vol. 112.
[ill HAROUN. M. A.: HOUSNER. ti. W. Seis
NO. I , I%.pp. Iin.
creasing their ability to dissipate seis mic design of liquidstorage rmkr Journal of
mic energy need to be examined. Technical Councils ASCE. Ncw Yark. Vol. IOi. 1231 BAKI'ON. D. C.: PARKER, J. V. Finite ele
The tank could either be anchored to No. 1,1981, pp. 191207. merit analyris "Jrhe sdvmic rmponre of mrhomd
its foundation with energy dissipating (121 VELETSOS, A. S Sci.smic r~.iponscund dr and tinanrhored liquid sioroge toilks Journal of
devices [29] or seismically isolated by sign of liquid sroragc tanks. Guidelines ior the
Earthquake Engineering and Structural Dynam
. is. NO. 3 . 1 9 ~ 7pp.
i c ~VOL . 299322.
special bearings [30.31]. Seismic Design of Oil and Gar Pipeline System5
A S E , New York, 1984 pp. 255370. 1241 PEEK, R.: JENNINGS, P. C. Simplifid
1131 VELETSOS, A. S.; TANG. Y. Soilsrrucrure on2alysi.s of ioiovchorcd ranks Journal of Earth
quake Engineering and Structural Dynamics,
References interaction effectsfor /akra//yexciiea liquidslor
Vol. 16.No. 7. 1988. pp. 10731085.
age r a n k Journal of Earthquake Engineering
[I]HANSON, R. D. ñehovior of liquidstorage and Structural Dynamics Vol. 19. No. 4, IYW, 1251 NATSIAVAS. S. Responre and failure of
tanks. the Great Almka Enrthquoke of 1964 pp. 4734%. Piridfilled t a n k under base rxcitation. PhD the
Proceedings of the National Academy of [14] VELETSOS, A. S el al. S<LFmic rrrponse of sis California Institute of Technology. Pasadena.
Scicnce. Washington. Vol. 7. 1973. pp. 331339. 1987.
anchored steel tanks Proceedings of the 'iñird
[2] U.S. Department oi Commerce. Earthquake Symposium on Curynt lssucs Related to N u  1261 LAU, D. T.; CLOUGH. R. W. Storic tilt be
damagr to water and swoge fudlities. Son Fernon clear Riwer Plant Structures. Equipment and havioitr of Amchorrd qlinddcal tanks. Report
do Earrhquoke of Fchrurrry 9. IY71. ProQedings Piping (Gupta, k K. Ed.). North Carolins State EERC 89.1 I . University of California. Berkeley.
of the National Oceanic and Atmospheric University, 1990. pp. 215, 1989.
Administration. Washington. Val. 2, 1973. pp. [IS]MALHOTRA. P. K.: VELETSOS. A. S. [27] WUNDERLICH, W.: SEILER. C. Nonlin
1351%. Seismic response of unonchrircd Ntd partially an car rrraiment of liquid filled .storage ranks under
131 GATES, W. E. Eleimrd ~,,dg,",,nd.~~pp~,rtchored liquidrromge 1ank.s. Report TR105MH. earthquake cxcirarion b.v II quarisrntic npproach.
rd rtrrl sroragr tanks. Recmnaissance repon, Electric Power Research Institute. Palo Alto, Advances in Computational Structural Mechan.
Imperial Couni,~California Earthquakc of Orto 1995. i s (Topping, B. H. V.. Ed.). Chapter 12.2. Saxe.
her 15. 1970 Earthquake Engineering Research [16] MALHOTRA. i? K. 8a.s~uplifring annly.sis Coburg Publications & CivilCump Presa, 1998.
Inrtitutc. Oakland. IYMI. of flexibly supported liquidsforage ianks Jour (281 Eurocode 8: Design pmvisiom of earth.
141 MANOS. G. C :CLOUGH. R. W. Tank dam nal of Earthquake Engineering and Structural quake resistance of SlruchtreS Part 4: Silos tonks
agp during the May IY8.I Coaling" Earthquake. Dynamic% Vol 24, No. 12. 1995. pp. iS911Ml7. and pipelines European Committee for Stan
Journal oi Earthquake Engineering and Struc 1171 MALHOTRA. P. K. Seismic r<cspoü.~of' dardization. B m s e I ~1998.
tural Dynamics. Vol. I . No. 4. 1985. pp. 4 4 M . soilsupported unanehorcd liqnidsmnige tanks [ZSl MALHOTRA, i? K. Seirmic strcngrhenuig
[SI HALL, J. F. (Ed.) Nrirthridge €arthquake of Journal of Structural Engineering, ASCE. New qf liquidsromge tanks with mergydissipating
Janiiary 17, IY94. rrctmtai.rsartce report. Vol. I. York. 1997, Vol. 123. No. 4. pp. 444450. anchorr Journal oi Structural Engineering. Vol.
Earthquake Engineering Research Institute, [is]WOZNIAK. R. S,MITCHELL, W. W. 80 124, No.4. IY%.pp.405414.
Oakland. 1995. sis of seismic design prosi.~ionrfor weldrd rtcrl [W] MALHOTRA, P K. New method fiir ,seis
161 BROWN. K. J. et al. Seismic performonce of oil storage tanks Sessions on Advances in Stor mic isolarion of liquidstorage t a n k Journal of
L a s Angeles w m r ranks Proceedings of the ageTank Design, American Petroleum Institute. Earthquake Engineering and Structural Dynam
Fourth US Conference on Lifeline Earthquake Washington. 1978. i vol. 26. NO. 8.1997, pp. ~391347.
Enpinerring (O'Rourke. M. J.. Ed.). ASCE. 1191 CLOUGH. R. W.: NIWA. A. Sriitic rilr test,s [3lJBACHMANN. H.: WENK. T. Sofiening in
New York. 199.5. pp. MX675. of a tall cylindrical liquidrroroge t m k . Report stead of strengthening for seismic rehahilitorion.
171 LUND. L. V. Lifeline utilities lessons, North EERC 7906. University oi California, Berke Structural Engineering Intcrnational. IABSE,
ridgr Earrhqrrakr. Proceedings of the Fourth ley, 1979. Zurich, Vol. 10. No. 1.2MX). pp. 6165,
U.S. Confcrencr on Lifeline Earthquake Engi
neering (ORourke. M. J., Ed.). ASCE. New
York. IYYS. pp. h76683.